I recreated a BMI algorithm based on the base paper, Choi et al. I compared the common spatial frequency subspace decomposition (CSFSD) method of Choi with the older common spatial patterns (CSP) used in Ramoser et al.
My team designed an experiment of reaching out to an object with the left or right hand, upon cue. The subjects’ EEG was recorded using the Emotiv EPOC BCI device. Parallelly, the EEG motor imagery dataset from the Physionet database was used to test out the algorithms.
Ramoser had a higher accuracy based on the simulations compared to Choi. This was because the CSFSD transforms had some runtime error due to their complexity.
A subject during data collection session. The subject here is wearing the Emotiv EPOC headset, which is recording his EEG. (This can be observed in the background of the screen.) On the screen, you can see the LabVIEW light sequence program that gives visual cue to the user in the form of a light and a written message at the bottom. The subject imagines and then touches the orange object placed in front of him upon cue. The LabVIEW program sends markers to the recorded EEG.